Michael Patkin's

   The Ergonomics of Natural Written Language

Publication history, Reflections & comments



Surgery & ergonomics


Information design

Editorials, book reviews



Abstract: This paper argues for re-directing ergonomic attention from muscle to mind, from brawn to brain. The argument parallels the change from Industrial Society to Information Society. The growth in volume and complexity of information means that information must be designed to fit human capacity and limitations, like the re-design of physical work in recent years.

The two main human limits to handling information are short term memory and "schemas" - scripts or mental templates according to which new information is .............................

processed. These two factors are the basis for clear and simple writing, and effective lay-out of text. To make text easier to read, such layout should include boxes of head-
line phrases


There are 7 million computer terminals in the United States, with annual sales of a further million. Many people cannot cope with difficult instructions for using them. Some have changed jobs to avoid computers, and others have been driven to sabotage. The problem is a man-machine mismatch. It is an ergonomic problem of handling information.

Ergonomics was not a worry for primitive man. He picked up sticks he could handle, and shaped flints till they fitted his palm. Once inside his cave, he built up the fire till he felt warm enough. Much later, the industrial revolution created a gap in time and 

space between designer and user of tools, so that they were often badly designed for use. By the middle of this century, this gap had become so bad that it had to be closed deliberately, using the feed-back loop of ergonomics. Language today
is too complex

Language was also probably a simple matter for early man. One can imagine that primitive language was a few facial gestures, shouts,and grunts. We can easily imagine gradual changes in this language, with more detail learned by each generation in turn, to give the complexity of languages today.

Today, however much of language has become too difficult for human beings to cope with outside their normal routine - in surroundings which are new for individuals, or complicated, or badly organised, . This is because of the heavy load on limited human perception, on learning ability, and on memory, especially short-term memory. As a result, mankind is floundering in the transition from Industrial Society to Information Society. However help is at hand. Solving these problems will be even more useful than ergonomic analysis of the physical workplace.

In the physical work-place of yesterday's Industrial Society, principles of Ergonomics have now been worked out in detail and made widely known. Ergonomics of handle design, controls, manual handling, safety, fatigue, pulse rates, respiration, and the rest - these have all been noted, scrutinised , dissected, analysed, investigated, measured, argued about, mulled over, discussed, presented, published, photocopied, recopied, read briefly, filed, stored, and - one suspects - largely not read or used nearly enough.

So much for traditional ergonomics. But in developed countries, it is now almost a decade since the number of office workers outstripped the combined numbers of farm and factory workers.


In today's Information Society, the new workplace is one that involves less back and leg muscle, joint, and bone, and more brain, eye, ear, hand, and voice. Of course,

is not enough
using keyboards and screens is still a largely physical problem; it involves applied anatomy, physiology, mechanics, though it also involves applied psychology, and industrial relations. And the current story of Office Ergonomics is that of

huge payments for repetition injury, big advertisements for new styles of furniture, and a growing number of instant experts in this area.

Another, bigger, more expensive and remorseless problem has developed more insidiously. It is the handling of large quantities of new information, and goes far beyond the older question of humans as information processors, handling dials and displays. NaturalThis is the problem of coping with textual Languageinformation.

A capacity for information handling is one of the distinctive features of Homo sapiens - perception, organisation, storage, communication, and decision-making. Obviously machines are faster and more accurate for detail, but brains are much better in some vital respects; these respects are size of memory, pattern recognition, and decision based on incomplete evidence. (Machines and people can also both be programmed to give untruthful responses, an appropriate comment for the year 1984).


Natural language is what a person is used to

Language is usually defined as the set of sounds, symbols, and gestures used by members of a group in communicating with one another (transmitting their recording of experience, including internal mental processes). It has been argued that language depends on words, and that animals and babies don't have language.

Clearly this is wrong in the sense that we can think and remember with our muscles, communicate with other body postures and movements, and imagine with symbols and concepts. A family's cat has no problem in "asking" to be let out the door. Many human problems are alleged to be due to erroneous body language, and the success of advertising is at a different, behavioural, level to the mainly cognitive problem discussed here. This analysis is aimed at problems of written informative language.

Language is natural when it is used easily and clearly on the basis of easy familiarity, and does not require the intervention of conscious analysis.

Examples of natural language are:

1. A five-year old child in Paris when he is speaking simple everyday French, perhaps also shrugging his shoulders, elevating his eyebrows, and extending his upturned hands to indicate "I don't know".

2. A surgeon discussing choledochoduodenostomy among a group of other surgeons, or writing about it in a specialist journal.

3. A horse-racing fan listening to a radio broadcast of results of races, spoken rapidly by the announcer.

4. An ergonomist, used to working with statistical data, reading or constructing chi-squared correlations.

Natural language is used by people when they are in a habitual environment. It is also important to turn the question around, and ask "what is the opposite of natural language - what is unnatural and difficult language"? Consider the factors which make it hard to understand text:

1. Content - unfamiliar concepts (or "schemas", to use this important expression of Immanuel Kant's in its new fashion, which is considered in detail later). Unfamiliar vocabulary - one person's natural language and abbreviations are another person's jargon.

2. Transmission - defects in source/route/destination, or rate. Such factors include mumbling, poor print quality, surrounding distractions, deafness or lack of reading-glasses etc., and information over-load.

3. Lack of background, of appropriate schemas; emotional blocks induced by fear, insecurity, bullying, resentment etc.

These three factors correspond nicely to a scheme of source/route/destination, the model for spread of infection. In this case the infection is ideas, not bacteria.

THE VITAL BASIS Naturallanguage

FOR NATURAL LANGUAGE relies on schemas

& short-term memory

The two factors which seem most important for dealing with written information are "schemas" and short-term memory.

Schemas are built-in mental check-lists, theories about the outside world. They are a back-ground of memory against which incoming messages are compared, and onto which they are grafted with varying degrees of efficiency. Reading and listening are much more than functions of the eye and ear, It is useful to think of reading in terms of a system, where two streams - written material and the reader's memory - converge to be processed in short-term memory, and then, perhaps, to enter long-term memory. (There is more to the reading mechanim, mentioned later). Natural


The second factor is short-term memory. Its limitations have been defined many times, most notably by Harvard psychologist George Miller in a paper called "The Magic Number Seven, Plus or Minus Two." An average person finds it hard to remember more than seven unrelated items, such as numbers, for more than a few seconds if he does not keep repeating them. Telephone numbers are familiar examples.

Let us study these two factors in more detail:


A schema is a set of rules in the mind. (There are many such sets). By means of these rules, or categories, images of the external world are shaped to fit the mind. The term "schema" was introduced by the German philosopher Immanuel Kant in 1781
Schemas are mental patterns for viewing the world

An older term "categories", had been written about by Aristotle, who presented 10 categories in which all parts of the external world could be classified. These included quality, substance, relations of size, position, and so on. Aristotle regarded these categories as absolute. Kant had a similar approach, but used "schemas" for the mind's way of thinking about and classifying things and ideas much more widely.

There have been two further changes in using the term "schema". The first, by the renowned psychologist Piaget, considered it to be a mental framework or outline, which refuses to be defined consciously (Penguin Dictionary of Psychology). The present trend is to consider a schema as an organised way of thinking about experience. Because it is such a handy concept, this is how "schema" will be considered here - a mental representation of experience, like a blue-print or check-list. A tourist, a geologist, and a property developer -each has a schema for the countryside different to that of other people.

A schema provides the mind with a unified picture of the outer world, rather than a scattered set of images. (It has a lot in common with the German term "gestalt"). A schema of a triangle does not correspond exactly to any triangle in the real world, but it allows the mind to put a real three-angled plane figure into the category of "triangle". There are similar schemas for bicycles, for the colour green, for a probable case of appendicitis, and other things the mind deals with.

Type of schema include objects, space, time, cause-and-effect, quantities. Kant described the schema of causality as linking two items, where one is the cause and the other the effect. Some primitive people may not link intercourse and pregnancy - such detailed schemas are learned, though some may be built-in, and other constructed by the imagination. Other examples of learned schemas are a mental picture of the structure of the atom, or what Napoleon may have looked like. Schemas are the "deep structure" of thought; they provide the ability to recognize patterns, to make sense out of things. (The term "script" has been used in a similar way by Schank, to describe a set of expectations, or codified information - see Kendig,1983).

A teacher or writer does best by using a reader's existing schema, or teaching a simple new one, which may be termed a "model". Journalists have long used Kipling's checklist:

"I keep six honest serving men,

(They taught me all I knew);

Their names are What, and Why, and When,

And How, and Where, and Who."

(From the Just So Verses).

Many activities have such checklists. Part of the work of a designer, or writer, is to use schemas which his client or reader will be familiar with. The direction of rotation of control knobs, for example, has a common stereotype relation to expected travel of a liner pointer, which varies with different types of displays.

Alphabetical arrangement of items is obvious. Less obvious are the rules for surnames starting with Mac and Mc. What about upper-case terms like TNT, and punctuation symbols? Computing has clear and simple rules for the order of all symbols used, based on the number value of each character in machine code. However the non-computerist will be unaware of this, and needs it explained, at risk of boring others. Modern text-books are full of appendices, with crash-courses for the tyro in statistics, or Boolean algebra, or anatomy, or some other topic which will be familiar to some and not others. In a world of growing complexity, the need for Ad Hoc will also grow.

SHORT-TERM MEMORY - remembering 7 or 8 things for 30 seconds

Short-term memory is what you use between looking up and dialing a telephone number. It is remembering Naturala handful of items for up to half a minute, or Language longer with rehearsal, either silent or aloud. Short-term memory is destroyed by mental distraction, either an experimentally designed alternative task, or an unwanted extra load during mental work of some kind.

The steps in memorising may be summarized as:

Perception (good lighting, clear print).

Sensory store, less than 2 seconds.

Short term memory, less than 30 seconds, easily distracted, maximum of about 7 items, and schema-related.

Long term memory - up to a life-time, needs practice, and is schema-related, e.g. mnemonics.

Typical examples of short-term memory failure are those caused by "trans-coding" - briefly learning some new code to which actions must be fitted. This might be a series of computer instructions translated from BASIC to DBASE for example - such instructions as DO instead of RUN, ACCEPT instead of INPUT, and so on.

However much more data can be remembered, if it is the form of larger and memorable "chunks", either as familiar big units, or with a familiar key.For example the number 14916253649 is difficult to remember, if the reader hasn't noticed (or
"Chunks" are bigger lumps of remembered material

been told, which is equally effective) that it is the cramped up squares of the first seven numbers, thus:

1 4 9 16 25 36 49

Short term memory is the basis for capturing material from the outside world (which includes the body outside the mind), via the sensory store. This immediately explains the importance of lay-out and Naturalspacing - data can be taken in in meaningful or Languagefamiliar chunks, which can then be subject to further chunking, or aggregation into still larger units.

Chunking is needed for learning most skills. Examples: 1. Reading of words letter by letter, by the small child, who has no brain store of word chunks against which to compare them. 2. Typing by the hunt-and-peck method. 3. Trying to ride a bicycle, or to juggle three balls, according to numbered instructions. Sometimes this has been called the "hierarchical" stage of skill acquisition.


Because of problems of jargon in the British Civil Service, Gowers wrote a short materpiece called "Plain Words". Every writer should have a copy of it.
Books by Gowers,Flesch, & others

In the United States, the "Fog index" was devised by Gunning, and a readability index by Flesch. These were measures of how simply, or otherwise, written language was put together. A recent study called "Chemistry textbooks - are they readable?" (Knutton, 1983) showed that many were not, using such formulae. Bad complicated writing is such a problem today that a presidential task force has looked at legal language, and military contracts in the United States now have readability requirements for manuals to accompany computer programs (MIL-STD-1472C). Previous editions of the same standard (1472B, 1978) had been confined to physical aspects of work-place and equipment design, an interesting mirror to the introductory paragraphs of this paper.

However such counts of the length of words and sentences tell only part of the story. Such formulas can be used to show that comic books are more difficult to read than writings by James Joyce, the author of "Ulysses", one of the hardest of modern famous books to understand.

More valid measurements of clear writing would permit long sentences whose structure was easily visible to the reader, and also allowed long words where these made understanding and remembering easier. Such writing would be based on schemas which are familiar, or which can be learned easily.


Text is not the only kind of written language, whether on paper or on the computer screen. Pictures and diagrams can be added very usefully - "one picture is worth a thousand words".
Natural Language

New computer systems like the Apple Lisa use diagrams called "icons" - a filing cabinet to represent a file, a waste-basket to represent a deletion, and so on. There can also be added beeps, warning lights, and even spoken messages (at present in a Japanese-American accent). The ergonomics of diagrams (e.g. line diagrams in 3-dimensions are generally best) is a separate and large topic not considered further here.


These mean:

1. Legibility

2. Layout

3. Language

4. Sentences

5. Paragraphs

6. Pages

7. The ergonomics of book design

These are the nuts-and-bolts of practice. A detailed paper on this was published in shortened form in a recent issue of the Australian Computing Bulletin. Only a brief outline is presented here.

1. LEGIBILITY. Typeface should be large enough (12 point), of a particular style (e.g Cheltenham, Times Roman etc., with serifs), and semi-bold rather than faint. Ordinary language has great redundancy, an. mis..ng ch.r.cte.s c.n be coped with. However technical or unfamiliar writing is another matter. It would be easy to confuse IO, the moon of Jupiter, with the decimal 10. In a san-serif style, it is hard to tell apart "Ill" from "111". Computing has many such problems. CAPITAL LETTERS, SOMETIMES USED FOR EMPHASIS, ARE ACTUALLY SLOWER TO READ, LACKING THE USUAL WORD-SHAPES OF LOWER-CASE.

Type and print-wheels should be clean, ribbons should be fresh, and paper should be of good contrast and not glossy. Photo-copying should be black-on-white, and not smudged grey-on-grey. Such rules should hardly need writing down, yet everyday experience is that they are ignored.

2. LAYOUT AND SPACING. Wordscanbecrampedtogether, or t h e y m a y b e s

p a c e d t o o m uc h. Lines should best have

35 to 55 characters. Longer ones are more difficult to read - more eye movements are needed, and the eye does not drop down easily to the start of the next line. Paragraphs, whose content is discussed below, should be shaped with a small indent at the start and clear lines above and below.

Large tables of figures should have several spaces between columns, and a blank line after every four or five lines of data. Numerals should be right-justified, on the left of text which is left-justified. The design of forms and of video text screens are specialties in their own right.

3. LANGUAGE. Words should be familiar ones, and not ambiguous, especially in instructions, when a word which is either noun or verb is confusing e.g. type. and copy. In spoken language there are clues from tone and timing, which are missing in text. Abbreviations also cause problems. Looking up what they mean is an example of "transcoding", discussed below. Abbreviations save time in some cases, and cost it heavily at other times. Now that the cost of computer memory is so much less, it should no longer be a constraint. While style is individual, easily read text has a good ratio of nouns, verbs, and adjectives.

4. SENTENCES. These may be only one word. Look! As units of information, they may over-load short-term memory. This occurs when the verb is hidden a long way from the start. How to write good sentences is a topic, not just a few brief remarks, but for books (see bibliography). Such books are weak on perceptual psychology; they do not discuss emphasis, or the failure to perceive words. Short ones, like "if", may be missed easily, UNLESS they are capitalised, or highlighted by placement at the start of a sentence. If this is done, other techniques are unnecessary (bold-face, underlining).

(Of course the books referred to were written in the days before cheap micro-computers allowed private individuals the luxury of word-processing and some limited type-setting facilities. This paper was written on an Osborne micro-computer, and revised many times).

Ergonomists have taken a special interest in the ergonomics of instructions, whose importance here is that they succeed or fail mostly as sentence structures. Much has been made in computing of "user-friendliness", where the problem is similar.

5. PARAGRAPHS should be self-contained units, Natural

relating to one item of informa Languagetion. The first sentence should summarise or indicate the content of what follows, to smooth the thinking of the reader, and to allow quicker preliminary skimming. The last sentence should, if possible, be a link to the content of the next paragraph.

In expository ("serious") writing, each single paragraph should, in general, be equivalent to a schema. The art, and science, of constructing paragraphs is the subject of an entire book by Roloff and Brosseit. Repetitive detail is often presented better as a table than as a stew of text. This clarifies relationships.

The shaping of paragraphs, discussed above, makes them easier to read.

6. PAGE DESIGN is a neglected subject. A big A4 size is best, with lots of white space, and grouping of material within it. Excellent discussion and examples of this are given by James Hartley in his book on "Designing Instructional Text". A page may be designed like a map in an atlas, as a self-contained unit of information.

7. BOOK DESIGN. It should be easy to identify a book by its cover, and spine, and to keep in order by having the Dewey number placed on the spine by the publisher. Book bindings should allow them to stay open without, for example, taking one hand from a key-board. Page numbering should not need mention, yet it often a problem with technical publications.

A book can be considered as a machine for the storage and retrieval of information. It should therefore be organised for efficient use, with adequate chapter divisions and names, index, glossary, context and style. Under-used devices include "bleeding" tags (blocks extending onto the paper margin for easier location of articles), and headings. "How to" books, especially computer manuals, too often neglect Whitfield's advice (1979): The aim of any procedural aid is "to aid the performance of specific tasks by reducing demands on memory and complex reasoning".

Recent advances in office equipment mean that "in-house" publishing will increase. What will be much slower in general use will be know-how for book-design along the lines just discussed.RULES FOR GOOD WRITING

These should be part of the stock-in-trade of any scientist, especially an ergonomist, quite apart from the subject of this paper. These rules are the subject of several excellent books (see Flesch, Gowers etc.). Here one merely recalls, in dense unfriendly text:

1. Use short familiar words.2. Use the active rather than the passive voice. 3. Avoid negatives, especially double negatives.

4. Use "not" instead of an apostrophe ("don't), or a prefix ("not edible" instead of "inedible").

5. Make sentences short, and cut paragraphs thin, like salami, to make them digestible.

6. Revise, and re-write. Word-processors eliminate excuses for this. Edit for format, style, lan guage, substance, and context.

7. Break these rules at the right time.

Here is an example of legal business jargon, originally printed in small type. (It is from Brockman's work, which has been drawn on heavily for this paper):

"The unqualified word "Insured" includes the Named Insured and also:

(1) any relative .... (2) ....any person

provided, the insurance with respect to any person or organisation other than the Named Insured does not apply:

It has been re-written:

" You're covered for any accident involving an auto you own, borrow, rent, or use as a temporary substitute. We'll also cover accidents during maintenance, loading, and unloading. And we'll cover your liability to the owner for damage to an auto you borrow or rent.

We'll also cover anyone else who uses a covered car with your permission provided it's used for the purpose you intended." [an example is described].

Such reader-based writing presents instances Naturalrather than general principles, a conversational Languagestyle, metaphor, contractions, and first or second person pronouns. It is organised round the reader's questions.


There is a lot of bad writing which undoubtably costs time, efficiency, and temper. (There is solid evidence for this, which has not been presented here). The simple rules for good writing, quoted above, are often ignored. Beyond these, psychology provides important new insights about human capacity, with implications for the presentation of written language.

These apply to the authors of manuals, papers presented at conferences and published in journals, to books, screen format design, and instructions of many kinds. Applying the principles described would save money, and perhaps ensure safety.

However the urgent always takes precedence over the important, and most items in modern work-life seem urgent. For example, few professional and technical people take time to learn to touch-type. How can they be encouraged to handle language, and information, well?

The possible strategies are: Rewards for good presentation, and penalties for bad ones, with provision of facilities, especially advice and word-processing, for easier improvement of skills. The rewards and penalties may be money, recognition, and perhaps most importantly, simple awareness. This paper is one such small step to inculcate awareness.

What is your response?

Acknowledgement: This paper could not have been written without the continuing help of Mr. Don Ransted, librarian at the Whyalla Campus of the South Australian Institute of Technology.


This paper follows on from:

Patkin, M. and Molyneux, A.R. (1983, August) "Software Ergonomics", Australian Computer Bulletin, 7,7, 11-15. This is a shortened version of a much more detailed manuscript. It was also based on lengthy discussions with the authors of the following two items:

Brockman, R. John. Seminars, "Writing Better Computer Software Documentation - A Short Course." 1983.

Galitz, Wilbert O. (1980) Human Factors in Office Automation. Life Office Management Association. (237 pp.)

Material most closely relevant to this paper was gleaned from:

Lindsay, P.H. and Norman, D.A. (1977) Human Information Processing: An Introduction to Psychology. 2nd edition, Academic Press.

Norman, D A. Memory and Attention. An introduction to human information processing. Wiley, 1976.

Roloff, Joan G. and Brosseit, Virginia (1979) "Paragraphs". Glencoe Publishing Co., Encino, California (239 pp).

Smith, Frank (1978) Understanding Reading: A Psycholinguistic Analysis of Reading and Learning to Read. 2nd Ed. Holt, Rinehart & Winston (263 pp).

Specific References

Broadbent, D.E. (1977, March) Language and ergonomics. Applied Ergonomics, 8.1, 15-18.

Card, Stuart K. (August 1978) Studies in the psychology of computer text editing, Xerox Palo Alto Research Center, SSL-78-1, San Jose, California.

Chapanis, A (1965) Words, Words, Words. Human Factors, 7, 1-17.

Christie, Linda Gail (1982, August) How to Evaluate Documentation Manuals. Interface Age, 58-64.

De Marco, Tom (1978) Structured Analysis and System Specification. Yourdon.

Durrett, J. and Trezona, J. (1982, April). How to Use Colour Displays Effectively. BYTE, 50-53.

Dwyer, Barry (1981, September) A User-Friendly Algorithm. Communications of the ACM, 556-561, reprinted in Larson.

Easterby, R.S. and Hakiel, S.R. (1981, September) Field testing of consumer safety signs: The comprehension of pictorially presented messages. Applied Ergonomics, 12.3, 143-152.

Flesch, Rudolph (1946) How to Write, Speak, and Think More Effectively. Signet.

Glencross, D. J. (1977) The Control of Skilled Movements. Psychol. Bull. 84, 14-29.

Gowers, Ernest (1968) The Complete Plain Words. Pelican.

Hartley, James (1978) Designing Instructional Text. NaturalKogan Page, London. 124 pages. Language

HMSO (1972) The Design of forms for Government Departments.

Kant, Immanuel (1781) The Critique of Pure Reason, translated T.K. Abbott, Encyclopaedia Britannica Inc. Chicago 1952.

Kendig, F. (1983) A Conversation with Roger Schank. Psychology Today, April, 28-36.

Larson, James A. (ed.) (1982) End User Facilities in the 1980's. Computer Society Press. 503 p.

Martin, James (1973) Design of Man-Computer Dialogues, Prentice Hall.

Miller, G.A. (1956) The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review. 63, 81 -97. (Reprinted in Norman's book, above, and in the Pelican book on memory).

Patkin, M (1971) Books As Machines. Medical Journal of Australia, 1, 44-45.

Patkin, M. (1981) Ergonomics of the Printed Information Explosion. Proceedings, 8th Annual Ira Symposium on New Vistas in Management and Quality of Work Life, 367-376, Tel Aviv.

Phillips, R.J. (1979, December) Why is lower case better? Some data from a search task. Applied Ergonomics, 10.4, 211-214.

Poulton, E.C. (1970) Ergonomics in Journal Design. Applied Ergonomics, 1, 4, 207-9.

Rutkowski, Chris (1982, October and November) An Introduction to the Human Applications Standard Computer Interface, Part 1, Theory and Principles. Byte, 7, 10, 291-310. Part 2, Implementing the HASCI Concept, 7, 11, 379-390.

Shneiderman, Ben (1980) Software Psychology. Winthrop.

Smith, D.C., Irby, C, Kimball, R, Verplank, B, and Harslem, E. (1982, April) Designing the Star User Interface. BYTE, 242-282.

Style Manual for authors, editors, and printers (1978). Australian Government Publishing Service, Canberra, 3rd edition.

Whitfield, David (1979) The Design of Procedural Aids, in Proceedings of the Sixth Annual Ira Symposium on Human Engineering and Quality of Work Life, Tel Aviv, 479-517.

Wright, Patricia (1981, September) "The instructions clearly state ...." Can't people read? Applied Ergonomics, 12.3, 131-141. (This paper includes a valuable and extensive bibliography).

Woudhuysen, James (1982, December) Software: Good Design. Microcomputer Printout, 3, 13, 33-35.


   The Ergonomics of Natural Written Language 
Michael Patkin
Surgeon, The Whyalla Hospital
A. Robert Molyneux
Programmer-Analyst,BHP Company Limited, Whyalla